IPv6 big packet blackhole in Teleglobe/TATA backbone?
Nuno Vieira
nuno.vieira at nfsi.pt
Tue Aug 19 20:24:36 CEST 2008
humm...
here goes a traceroute from here.
traceroute to poy.remlab.net (2001:41d0:1:a0d6::401:1983) from
2001:b18::211:43ff:fefd:90f6, 30 hops max, 16 byte packets
1 ge-1.Edge1.ip6.Lisbon.NFSi.pt (2001:b18::ffff) 0.229 ms 0.205 ms
0.111 ms
2 2001:5a0:1500::1 (2001:5a0:1500::1) 31.96 ms 1.723 ms 0.613 ms
3 2001:5a0:2a00:100::5 (2001:5a0:2a00:100::5) 13.349 ms 13.094 ms
13.23 ms
4 2001:5a0:1a00::1d (2001:5a0:1a00::1d) 32.589 ms 32.583 ms 32.593 ms
5 2001:5a0:1a00::3e (2001:5a0:1a00::3e) 32.34 ms 32.459 ms 32.469 ms
6 2001:5a0:1900::a (2001:5a0:1900::a) 32.838 ms * 36.727 ms
7 * 2001:41d0::591 (2001:41d0::591) 36.362 ms *
8 2001:41d0:1:a0d6::401:1983 (2001:41d0:1:a0d6::401:1983) 36.24 ms
36.197 ms 36.212 ms
Pekka Savola wrote:
> I and Remi have been trying to figure out an IPv6 big packet blackhole
> that seems to be occurring in Teleglobe network, specifically around IP
> addresses 2001:5a0:200::5 and 2001:5a0:200::10. I suspect MTU is not
> properly configured.
>
> This can be seen by a big packet ping or "http://www.remlab.net" if the
> traffic seems to hit the link in Teleglobe network.
>
> I've tried <IPNOC at tatacommunications.com>, but they keep asking for
> circuit ID and whether I'm their customer.
>
> Is anyone with Teleglobe/TATA on the list, or know a better contact?
>
> [Btw: interesting thing is that Telia seems to be exposing 1480B MTU in
> at least some parts of their backbone, but at least they're generating
> ICMPv6 packet too bigs and PMTUD has a chance of working..]
>
> ---------- Forwarded message ----------
> Date: Mon, 18 Aug 2008 18:55:21 +0300
> From: Rémi Denis-Courmont <rdenis at simphalempin.com>
> To: Pekka Savola <pekkas at netcore.fi>
> Subject: Re: remlab.net v6 PMTU problem?
>
> Sorry for the delay, I overlooked this email.
>
> Le mardi 22 juillet 2008 12:45:00 Pekka Savola, vous avez écrit :
>> Maybe I could raise this problem on the ipv6-ops list.
>>
>> Could you send the full traceroute/tracepath output from your end, and
>> confirm up to which router ping6 with 1500 sized packets works, and
>> where it fails?
>
> With not too big packets:
>
> % rltraceroute6 -I csc0-ws.ipv6.funet.fi 1000
> traceroute to 2001:708:0:f000::7100:2 (2001:708:0:f000::7100:2) from
> 2001:41d0:1:a0d6::401:1983, 30 hops max, 1000
> byte packets
> 1 * 2001:41d0:1:a0ff:ff:ff:ff:ff (2001:41d0:1:a0ff:ff:ff:ff:ff)
> 1.717 ms
> *
> 2 2001:41d0::592 (2001:41d0::592) 6.330 ms * 5.331 ms
> 3 2001:5a0:1900::9 (2001:5a0:1900::9) 4.869 ms 4.754 ms 4.686 ms
> 4 2001:5a0:1900::2 (2001:5a0:1900::2) 4.669 ms 4.576 ms 4.675 ms
> 5 2001:5a0:0:100::a (2001:5a0:0:100::a) 10.508 ms 10.328 ms
> 10.316 ms
> 6 if-6-0.core1.b1d-brussels.ipv6.teleglobe.net (2001:5a0:1400::1)
> 10.553
> ms
> 10.451 ms 10.400 ms
> 7 if-6-0.core2.ad1-amsterdam.ipv6.teleglobe.net (2001:5a0:200:200::1)
> 216.493 ms 94.318 ms 49.794 ms
> 8 if-0-0.core1.ad1-amsterdam.ipv6.teleglobe.net (2001:5a0:200:100::1)
> 18.189 ms 18.332 ms 18.095 ms
> 9 2001:5a0:200::5 (2001:5a0:200::5) 18.232 ms 18.247 ms 18.279 ms
> 10 2001:5a0:200::10 (2001:5a0:200::10) 19.794 ms 19.355 ms 19.623 ms
> 11 ldn-ipv6-b1-link.ipv6.telia.net (2001:2000:3010:1::1) 29.447 ms
> 29.534 ms 29.316 ms
> 12 ldn-ipv6-b1-link.ipv6.telia.net (2001:2000:3010::1) 60.893 ms
> 61.063 ms 60.760 ms
> 13 2001:2000:3080:d::2 (2001:2000:3080:d::2) 45.984 ms 45.987 ms
> 45.714 ms
> 14 helsinki0-rtr.funet.fi (2001:948:0:f035::2) 56.706 ms 56.023 ms
> 56.185 ms
> 15 csc0-x0000-helsinki0.ipv6.funet.fi (2001:708:0:f000::1:2) 52.700 ms
> 52.646 ms 52.587 ms
> 16 csc3-g0000-csc0.ipv6.funet.fi (2001:708:0:f000::30:2) 53.087 ms
> 53.001 ms 52.810 ms
> 17 2001:708:0:f000::7100:2 (2001:708:0:f000::7100:2) 52.578 ms 56.028
> ms 56.005 ms
>
> With MTU packets:
> % rltraceroute6 -I csc0-ws.ipv6.funet.fi 1460
> traceroute to 2001:708:0:f000::7100:2 (2001:708:0:f000::7100:2) from
> 2001:41d0:1:a0d6::401:1983, 30 hops max, 1460
> byte packets
> 1 2001:41d0:1:a0ff:ff:ff:ff:ff (2001:41d0:1:a0ff:ff:ff:ff:ff) 2.004
> ms *
> 2.275 ms
> 2 2001:41d0::592 (2001:41d0::592) 5.445 ms * 11.539 ms
> 3 2001:5a0:1900::9 (2001:5a0:1900::9) 5.044 ms 4.888 ms 5.176 ms
> 4 2001:5a0:1900::2 (2001:5a0:1900::2) 4.764 ms 4.710 ms 4.745 ms
> 5 2001:5a0:0:100::a (2001:5a0:0:100::a) 10.437 ms * 10.568 ms
> 6 if-6-0.core1.b1d-brussels.ipv6.teleglobe.net (2001:5a0:1400::1)
> 10.563
> ms
> * 10.720 ms
> 7 if-6-0.core2.ad1-amsterdam.ipv6.teleglobe.net (2001:5a0:200:200::1)
> 18.265 ms * 18.143 ms
> 8 if-0-0.core1.ad1-amsterdam.ipv6.teleglobe.net (2001:5a0:200:100::1)
> 18.783 ms * 18.122 ms
> 9 2001:5a0:200::5 (2001:5a0:200::5) 186.237 ms 215.084 ms 212.344 ms
> 10 * * *
> 11 * * *
> 12 * * *
> 13 * * *
> 14 * * *
> 15 * * *
> 16 * * *
> 17 * * *
> 18 * * *
>
> As for PMTUD:
>
> % tracepath6 csc0-ws.ipv6.funet.fi -n
> 1?: [LOCALHOST] pmtu 1500
> 1: 2001:41d0:1:a0ff:ff:ff:ff:ff 4.198ms
> 2: 2001:41d0::592 5. 8ms
> 3: 2001:5a0:1900::9 4.995ms
> 4: 2001:5a0:1900::2 4.733ms
> 5: 2001:5a0:0:100::a 10.408ms
> 6: 2001:5a0:1400::1 10.532ms
> 7: 2001:5a0:200:200::1 18.147ms
> 8: 2001:5a0:200:100::1 18.193ms
> 9: 2001:5a0:200::5 18.281ms
> 10: no reply
> 11: no reply
> 12: no reply
> 13: no reply
> 14: no reply
> 15: no reply
> 16: no reply
> 17: no reply
> 18: no reply
>
> I got an interesting response here:
>
> % ping6 -n csc0-ws.ipv6.funet.fi -s 1436
> PING csc0-ws.ipv6.funet.fi(2001:708:0:f000::7100:2) 1436 data bytes
>> From 2001:2000:3010::1 icmp_seq=1 Packet too big: mtu=1480
> 1444 bytes from 2001:708:0:f000::7100:2: icmp_seq=2 ttl=54 time=58.3 ms
> 1444 bytes from 2001:708:0:f000::7100:2: icmp_seq=3 ttl=54 time=59.7 ms
>
> --- csc0-ws.ipv6.funet.fi ping statistics ---
> 3 packets transmitted, 2 received, +1 errors, 33% packet loss, time 2001ms
> rtt min/avg/max/mdev = 58.321/59.019/59.717/0.698 ms
>
> After lots of pings (and resets of PMTU cache), it seems that:
> - packets over 40+8+1438 bytes get lost,
> - packets over 40+8+1432 bytes get a Too Big error from Telia,
> - other packets go through
>
More information about the ipv6-ops
mailing list